SDE 2 at Amazon Just Walk Out · Seattle, WA

Eliseo Nathaniel Ruiz Nowell

I am a software engineer with 5 years of full-time industry experience, and I take machine learning out of research notebooks and into production. Today I lead releases of the computer vision and neural network algorithms behind Amazon Just Walk Out — the technology that lets shoppers pick up what they need and leave without stopping to check out.

Portrait of Eliseo Nathaniel Ruiz Nowell
Amazon logo Just Walk Out

5 yrs

Full-time industry experience

200

Stores running an algorithm I shipped

50%

Receipt latency removed by a re-architecture

6

Co-op terms across 24 months

About me

Machine learning is only useful once it survives production.

I am a University of Waterloo Computer Engineering graduate who took the Artificial Intelligence option because I wanted to build systems that learn. Five years of industry work since then have made me an expert in everything that happens after a model is accurate — latency budgets, memory ceilings, offline evaluations, and the careful milestone plans that move research code into a mainline hundreds of stores depend on — and I have a track record of landing exactly that kind of launch.

That is the work I do now as an SDE 2 on Amazon Just Walk Out. I have been release lead on three algorithm launches — a streaming re-architecture, a Transformer-based beverage model, and a neural network that replaced a random forest classifier — each one coordinated across several engineers and researchers, and each one measured in latency, accuracy, and dollars saved.

Before Just Walk Out I worked on AWS X-Ray and OpenTelemetry, where most of my code was open source and reviewed in public. Six co-op terms before that took me through Newfront Insurance, Electronic Arts, Splunk, Yahoo!, and Rave — front end, game tooling, iOS, and backend in turn. I am happiest on the part of a problem where the measurement and the engineering meet.

Currently

Shipping computer vision to retail stores at Just Walk Out

Education

B.A.Sc. Computer Engineering, University of Waterloo

Based in

Seattle, Washington

Languages

English, Spanish and French (fluent), Mandarin Chinese (advanced)

Certification

AWS Certified Cloud Practitioner

Experience

Five years shipping, plus six co-op terms getting there.

Release lead on three Just Walk Out algorithm launches, an AWS open source maintainer before that, and a co-op run through insurance tech, game tooling, iOS, and backend services.

Amazon Just Walk Out

Aug 2022 – Present

Software Development Engineer II

  • Java
  • Python
  • Computer Vision
  • CNNs
  • Redis
  • RxJava
  • DynamoDB
  • SQL
  • Release lead, streaming receipt latency. Designed a streaming store orchestrator on Redis streams and Java RxFlowable that replaced 30 second batches with three parallel request/response pipelines, cutting receipt latency by 50% while coordinating milestones for four engineers.
  • Release lead, Transformer-based algorithm. Guided four SDEs and two researchers through a four milestone plan that brought 6,000 lines of research code for a beverage encoder/decoder model into production mainline, deployed to 200 stores after accuracy evaluations on more than 10 fixture types.
  • Release lead, neural network backend. Replaced a random forest classifier with a neural network across 171 stores for a 68% latency improvement and 25% fewer false positives, then used memray to cut peak on-premises memory by 15%.
  • Retired the legacy system for a saving of $4,000 per month and a week of developer effort each month, and called research code in-line from Java through JEP.
  • Shipped a sidecar service that automated account discovery and permissions, adopted by seven teams for allow-listing during 171 store launches, and removed overly aggressive DynamoDB locking to reduce store operation costs by 10%.

Amazon AWS X-Ray

Jun 2021 – Aug 2022

Software Development Engineer I

  • Python
  • OpenTelemetry
  • Java
  • Go
  • AWS
  • Docker
  • Developed multi-threaded, long-running performance tests now used across six GitHub repositories.
  • Reworked release CI to support ARM architecture for the popular OpenTelemetry Lambda layers.
  • Wrote technical posts after production launches teaching users advanced telemetry concepts, and merged fixes upstream in an open source library to unblock team CI and reduce code debt.
  • Commended by clients and team for meticulous handling of on-call responsibilities.

Amazon

Sep 2020 – Dec 2020

Software Engineer Intern

  • Python
  • OpenTelemetry
  • X-Ray
  • Docker
  • GitHub Actions
  • Split the OpenTelemetry Python project into core and contrib repositories over 90+ pull requests.
  • Released the Python package that makes OpenTelemetry SDK traces compatible with AWS X-Ray.
  • Built Docker sample apps showcasing OpenTelemetry and AWS compatibility, and added automated benchmarking to the GitHub Actions workflow.
  • Presented proofs of concept to OpenTelemetry Python SIG stakeholders and internally to Amazon engineers.

Newfront Insurance

Sep 2019 – Dec 2019

Full Stack Intern

  • TypeScript
  • node.js
  • React
  • Postgres
  • Datadog
  • CircleCI
  • Added authentication and authorization to every endpoint, together with a redesigned integration test suite.
  • Owned front end changes in React and node.js that improved backend performance, and learned cookie security to enable SSL validation on all website traffic.
  • Ran Salesforce and Postgres SQL queries to repair data integrity issues quickly.
  • Built a Datadog metrics dashboard whose tagging pinpointed a node website bottleneck.
Show four earlier co-op terms Hide earlier co-op terms

Electronic Arts

Co-op term

.NET Developer Intern

  • C#
  • WPF
  • XAML
  • C++
  • Frostbite
  • Maya
  • Led an audio mixer tool UX refactor and optimized the C# datasource to UI synchronizer.
  • Completed a level asset .fbx exporter for the Frostbite engine and linked it to Maya as a UI extension that artists praised.
  • Wrote the C++ audio mixer serializer used in the Frostbite game pipeline build routine.

Splunk

Co-op term

iOS & Backend Developer Intern

  • Swift
  • RealmDB
  • Python
  • twisted
  • AWS
  • Engineered a "CSS-like" iOS layout library that improved UI development time.
  • Integrated local storage with RealmDB using a thread safe async multithreading design, and surfaced Firebase remote notifications through the observer pattern.
  • Wrote async Python handlers and combined a key-value store with the Splunk API to serve dashboard CRUD requests.

Yahoo!

Co-op term

Backend Developer Intern

  • node.js
  • MySQL
  • express
  • AWS
  • Mocha
  • Memcached
  • API and database developer for the Yahoo! View product, writing node.js CRUD handlers over MySQL.
  • Used AWS CloudSearch faceting with Elastic Beanstalk and SQS to queue async server requests over HTTPS.
  • Finished an episode stream SQL fetch tool with retries and genre/offset query parameters, and validated assets with YAML Swagger files plus Mocha system and unit tests.

Rave

Co-op term

iOS Developer Intern

  • Swift
  • Objective-C
  • Xcode
  • CocoaPods
  • Fixed bugs and crashes and revamped the UI in a Swift and Objective-C codebase, optimizing performance with multi-threaded async requests.
  • Localized the app into four languages and refactored the user flow across every screen with a UX designer.
  • Polished git skills resolving conflicts, rebasing, and cherry-picking commits.

Toolkit

What I reach for.

Languages

  • Java
  • Python
  • JavaScript
  • TypeScript
  • SQL
  • C#
  • C/C++
  • Swift
  • Objective-C

Platform & data

  • AWS
  • DynamoDB
  • Redis
  • MySQL
  • Postgres
  • Docker
  • Linux
  • git
  • node.js

Machine learning

  • Computer Vision
  • CNNs
  • Transformers
  • NumPy
  • Jupyter
  • MATLAB
  • memray
  • JEP

Frameworks & tooling

  • OpenTelemetry
  • RxJava
  • React
  • iOS
  • Android
  • WPF
  • Mocha
  • GitHub Actions
  • Datadog

Projects

Things I built because I wanted them to exist.

Color Conquest running on an iPhone

Color Conquest

  • Swift
  • Bluetooth
  • AdMob

A multiplayer iOS strategy game. I designed a Bluetooth connected state machine for cross-device communication, populated tables and collections from dynamically changing data, and tuned the end-of-game algorithm so slower devices stayed responsive.

WAT 2048 game screen

WAT 2048

  • Java
  • Android
  • Sensors

An Android take on 2048 driven by accelerometer and rotation sensors. I replicated the game with object oriented Java, coordinated the team as delegated project manager, and profiled the running game loop for cross-OS performance.

Pray4Me backend logo

Pray4Me Backend

  • node.js
  • express
  • MongoDB

A server for a community prayer app. Express handles CRUD requests, a driver stores entries by authenticated Facebook id in MongoDB, and authorization tokens against bibles.org power verse lookup alongside a separate image upload path.

Education

Waterloo, with the artificial intelligence option.

University of Waterloo

B.A.Sc. Computer Engineering

Option
Artificial Intelligence
Cumulative average
90.3% over 8 terms
Honours
6× Term Dean's List
Co-op
6 terms, 24 months

Selected coursework

  • Distributed SystemsECE 454
  • Fundamentals of Computer IntelligenceECE 457b
  • Probabilistic ReasoningECE 493
  • Computer VisionCS 484
  • Programming for PerformanceECE 459
  • Algorithms & Data StructuresECE 250

Beyond the keyboard

Interests

  • Arduino
  • Half marathons
  • Chinese calligraphy
  • Level 9 RCM piano

Volunteering

  • Church app developer
  • Worship team pianist
  • ECE mentor

Contact

Let's talk about the hard part of your problem.

Whether it is a model that needs to survive production, telemetry that needs to make sense, or a launch that needs a plan, I would be glad to hear about it.